In one flat I lived in, the power bill was never late, and that is not the compliment it sounds like. It was never late because the same flatmate paid it, in full, three months running, and spent the rest of each month waiting to be paid back by people who had stopped noticing the bill existed. Every flat I have lived in had a version of that person. They chased the rent. They remembered when the power was due. They were the only one who actually knew whether the bond had been lodged with Tenancy Services or was sitting in a landlord’s account somewhere. Everyone else assumed it was sorted.
It was usually not sorted. It was a spreadsheet somebody started in February and nobody opened after April, or a group chat where the message asking for the internet money scrolled away unanswered. The money in a flat is not complicated. Rent divided by four is not hard arithmetic. What is hard is everything around the arithmetic, and for a long time I had the wrong idea about what that was.
The case for a better tool
The obvious reading is that this is a tracking problem, and a tracking problem wants a better tool. There are plenty. A shared spreadsheet, a splitting app, a whiteboard on the fridge, a running tally in the group chat. Pick one, keep it updated, and the flat runs itself.
Except every flat I have described already had one of those, and every one of them still had the person who chases the rent. The tool was never actually missing. So a better version of the tool was never going to be the fix, and the first thing I built, a cleaner ledger of who owed what, solved a problem the flat did not really have. I know this the honest way: I built it, and the person it was supposed to save would have gone on doing the same job with a nicer screen.
The expensive act
I had misread the job. I thought the load on that one flatmate was record-keeping, the work of knowing who owes what. It is not. The record is the easy part. The load is the asking.
Someone still has to look at the ledger, notice that you are forty dollars short this fortnight, and decide whether to say something. That is the expensive act, and no ledger performs it. You can make the numbers perfectly clear and the hardest moment is still the same one: a person, at the kitchen table, choosing between letting it slide again or being the flatmate who brings up money.
Once I saw that, the whole category of bill-splitting apps looked slightly beside the point. They automate the calculation and leave the confrontation exactly where it was. They tell the flat what everyone owes, and then they hand the job of collecting it back to the one person who was already doing it. The spreadsheet gets nicer. The treasurer stays unpaid.
What the register cannot see
The bond shows how deep the checking problem goes, because the bond is the one number with an official answer that nobody in the flat can casually reach. A landlord who takes a bond must lodge it with Tenancy Services within 23 working days, and faces penalties if they do not. (Tenancy Services 2026) The failure is real enough to be prosecuted; Tenancy Services publishes the cases, including a landlord ordered to pay $38,713.76 over unlodged bonds. (Tenancy Services 2026) And the register itself has a blind spot you could not design better if you tried: it has recorded every lodged bond since January 1993, around $900 million of them held in trust, and by construction it cannot count the bonds that were never lodged at all. (Tenancy Services 2026) The system that holds the truth only holds the truths that arrived.
So the flatmate lying awake wondering whether the bond made it is not being neurotic. They are the flat’s entire verification department, staffed by whoever happens to mind the most, working without access to the records.
Let the screen say it
The move that removes the asking is small and slightly strange: make the money side verifiable without a person in the loop. Connect, read-only, to the accounts through open banking, and let the app confirm that a payment matching the rent actually landed in the right account on the right date. Not move money, not see a balance, not control anything. Just verify, and then say so on a screen everyone can see. When the screen is the thing reporting that rent is short, nobody at the table has to be. The truth becomes ambient, coming from the bank rather than from the mouth of the person everyone quietly resents, and the awkward conversation stops being anyone’s job because it stops being a conversation at all.
There is a reason it has to be read-only, and it is the same reason. A flat will never hand an app the power to move its money, and it should not. But it will let something look, and confirm what already happened. The narrow permission is not a limitation to apologise for. It is the exact thing that makes the app usable by people who do not entirely trust each other with money, which is most flats, most of the time. Verification is acceptable where control never would be.
Most shared obligations work like this somewhere; a flat is just where I learned it. At flat scale the fix stays small and specific. The app never needs to move a dollar. It needs to say, where everyone can see it, that the dollar moved. A better ledger upgrades the treasurer’s equipment; this retires the position. The treasurer was never really keeping books. They were standing in for a fact nobody could check, and once the fact can speak for itself, the role has nothing left to do. The flat gets back to being a flat.